您目前所在的位置:首页 - 期刊简介 - 详细页面

铁道科学与工程学报

JOURNAL OF RAILWAY SCIENCE AND ENGINEERING

第11卷    第4期    总第60期    2014年8月

[PDF全文下载]    [Flash在线阅读]

    

文章编号:1672-7029(2014)04-0109-06
基于径向基神经网络的铁路客货运量预测研究
吴华稳1,甄津2,王宇3,王富章1

(1.中国铁道科学研究院 电子计算技术研究所,北京 100081;
2.铁道部 信息技术中心,北京 100844;
3.中国铁路总公司 计划统计部,北京 100844
)

摘 要: 根据径向基神经网络具有分析非线性动态系统的混沌特性的特点,对铁路客货运发送量相关时间序列进行分析和研究,在Takens相空间重构的基础上,利用互信息方法求嵌入时延、伪邻域方法求嵌入维数;应用G-P方法和最大Lyapunov指数方法对铁路客货运量时间序列进行混沌识别;根据RBF神经网络的学习算法和辨识原理,对铁路客货运量预测流程进行分析。应用径向基神经网络对铁路客货运量自1999-01-01~2012-08-27共4 988 d的发送量为基础进行径向基神经网络预测;并对预测误差进行检验及对预测结果进行分析。研究结果表明:基于径向基神经网络预测值能很好地与实际值相吻合,因而在铁路客货运量相关时间序列中预测有广泛的实用价值。

 

关键字: 铁路运输; RBF神经网络; 相空间重构; 客货运量混沌时间序列; 混沌预测

Railway passenger and freight prediction based on RBF neural network theory
WU Huawen1, ZHEN Jin2, WANG Yu3, WANG Fuzhang1

1. Institute of Computing Technology, China Academy of Railway Sciences, Beijing 100081, China;
2. Information Technology Center, Ministry of Railway, Beijing 100844, China;
3. The Plan Department of Statistic, China Railway Corporation, Beijing 100844, China

Abstract:Based on characteristic of RBF neural network theory to analyze the chaos of nonlinear dynamic systems, railway passenger and freight volume time series were analyzed. Specifically, on the basis of Takens phase space reconstruction, firstly mutual information method was used to calculate embedded time-delay and false neighbor method was utilized to calculate embedded dimension. Then G-P method and maximal Lyapunov index method were adopted to identify the chaos of railway passenger and freight volume time series. The next step, the prediction course of railway passenger and freight volume was analyzed using the learning algorithm and the identification principle. Finally RBF neural network theory was used to conduct prediction on railway passenger and freight volume from January 1st 1999 to August 27th 2012 with a total of 4988 days. Thereinto the prediction error was examined and the prediction result was analyzed. The result shows that the predicted data using RBF neural network theory is in accordance with the real data. Therefore RBF neural network theory has extensive and practical value in railway passenger and freight volume time series prediction.

 

Key words: railway transportation; RBF neural network theory; phase space reconstruction; passenger and freight traffic chaos time series; chaos prediction

ISSN 1672-7029
CN 43-1423/U

主管:中华人民共和国教育部 主办:中南大学 中国铁道学会 承办:中南大学
湘ICP备09001153号 版权所有:《铁道科学与工程学报》编辑部
------------------------------------------------------------------------------------------
地 址:湖南省长沙市韶山南路22号 邮编:410075
电 话:0731-82655133,82656174   传真:0731-82655133   电子邮箱:jrse@mail.csu.edu.cn